AMBER Archive (2002)

Subject: allocation memory error

From: Ioana Cozmuta (ioana_at_nas.nasa.gov)
Date: Fri Oct 18 2002 - 14:50:22 CDT

Hi,

I'm getting the following error when running a minimization in sander.

| Flags: SGIFFT MEM_ALLOC MPI RLE ROWAT HAS_FTN_ERFC
| NONPERIODIC ntb=0 and igb=0: Setting up nonperiodic simulation
| New format PARM file beingparsed.
| Version = 1.000 Date = 10/07/02 Time = 12:55:44
 NATOM = 34759 NTYPES = 16 NBONH = 18344 MBONA = 16779
 NTHETH = 36099 MTHETA = 22715 NPHIH = 68670 MPHIA = 41972
 NHPARM = 0 NPARM = 0 NNB = 181415 NRES = 2869
 NBONA = 16779 NTHETA = 22715 NPHIA = 41972 NUMBND = 43
 NUMANG = 88 NPTRA = 40 NATYP = 29 NPHB = 1
 IFBOX = 0 NMXRS = 24 IFCAP = 0 NEXTRA = 0

 Failed to allocate memory for ipairs: 604076661

The last suggestion was to find a cluster with at least 600MB of physical
memory. I am running this job on an origin cluster with 200MB memory per
CPU. I've tried to run this job on increased number of processors from 8
to 128 but it does not seem to make any difference.
>From what I understand looking into the code is that ipairs takes values
from 1 to lastpr which is equal to MAXPR. In sizes.h there is the
following part that allocates a value to MAXPR

#ifndef MEM_ALLOC
   parameter (MAXREA=2000000)
   parameter (MAXINT=2000000)
   parameter (MAXHOL=400000)
   parameter (MAXPR=5400000)
#endif

I understand that the above value for MAXPR is only used if the
-DMEM_ALLOC flag is not set in the MACHINE file and that actually what
happens is that these sizes are determined by the code and allocated at
run-time.

My questions are:
1. Why is the program not able to allocate the right amount of memory? I
mean when I run on 128 processors the available physical memory is
25 Gbytes. Is it possible that this is still not enough?

2. Would it make any difference if I would compile the program with a
larger value for MAXPR?

I would appreciate if anyone could help me understand how this things are
related and how they work and could suggest anything to help me solve this
problem.

Thank you,
Ioana